A slew of recent studies about meat and health has set off a veritable firestorm this fall, with a committee of physicians even filing a federal petition against the Annals of Internal Medicine, the journal in which the most recent study was published.
Why are so many people so angry?
After examining the literature on meat and health outcomes, the authors of that study issued new guidelines saying that we are, in general, fine to continue with our current levels of meat consumption. This recommendation flies in the face of oft-cited research on the dangers of meat consumption and of current nutritional guidelines.
The solution to the public trust problem is not to refrain from publishing studies because they run counter to the current narrative.
More than a few experts are upset about this. For many, the controversy is not just about meat. There is a serious concern that all this research whiplash is eroding public trust in scientific and medical experts. We’ve been here several times before. Eggs, for instance, were bad for us, then fine.
Now we’re doing the same thing with meat. The public is frustrated by nutritional guidelines constantly being in flux, with clashing opinions from experts. However, the solution to the public trust problem is not to refrain from publishing studies because they run counter to the current narrative. The solution is to address the problems in the field of nutritional epidemiology, to produce reliable data on which we can confidently base our nutritional recommendations.
Nutritional research is difficult because getting people to eat one way or another for a long period of time is quite a feat and, considering that disease is the outcome being examined, potentially unethical. Consider the implications of asking people to eat or not eat a specific food for research purposes, when your prediction is that the foods being assigned either cause or prevent disease. So for these and other reasons, nutritional studies are largely observational, meaning that we just ask people what they eat and correlate that with disease.
We also rely on food questionnaires that are often administered only a scant number of times over years. Problems abound with this methodology. People’s diets can vary greatly over time, their memories are often unreliable, and participants may alter their answers when they know a medical or research professional will be examining them.
Beyond these factors, which can only be statistically controlled for, to a certain extent, the data sets that result are still far from ideal. An incredible number of variables, including participants’ levels of physical activity and socioeconomic factors like income and education level, can be related to what people eat. These factors may be more important than specific nutrients when it comes to health outcomes.
Additionally, the sheer number of foods and nutrients (eggs, steak, lunch meat, protein, carbs, fat, saturated fat, sugar, etc.) and health outcomes (cardiovascular disease, stroke, diabetes, many types of cancer) stretches far past the ability of statistical testing. This adds up to a lot of noise that isn’t always helpful.
These problems are further compounded by the fact that the size of these data sets allows for findings that are so small that they don’t matter to individuals. They’re statistically significant, but not clinically significant.
Consider this: A 2014 study examining the third U.S. National Health and Nutrition Examination (NHANES III) data set reported that high protein intake from nonplant-based sources was associated with increased risk of overall mortality in middle-age individuals. A study published the previous year, using the exact same data set, found no significant association between meat consumption and mortality.
In other words, observational data can only take us so far. Some argue that because observational data form the basis for guidelines on smoking, they can do the same for guidelines on nutrition. But here’s the rub, the odds ratios in smoking studies are miles apart from the odds ratios in nutrition studies. (An odds ratio is a statistic that describes the association between an exposure and an outcome. A ratio of 1 means that an exposure has no relationship to an outcome.) The odds ratios in meat and health studies often hover around 1. — The odds ratios for cigarette smoking and lung cancer range from 16.8 to 103.5. So the danger from smoking and the danger from meat aren’t comparable.
It is important to note the real message here, which is not that you should eat meat with wild abandon, or that meat is suddenly a health food.
With the actual risks so small, it is reasonable to consider the risk versus the joy derived from eating meat. The studies in question did in fact assess data on meat-related values and preferences, though this has also been a point of contention for some experts.
Though not assessed in the current studies, there are other, more valid reasons to eat less meat, such as concerns revolving around animal welfare and climate change. We know these are valid because we have good data to tell those stories.
It is important to note the real message here, which is not that you should eat meat with wild abandon, or that meat is suddenly a health food. The message is that the evidence we have linking meat consumption to disease outcomes is poor. It’s probably too poor to make a recommendation that people avoid it completely or reduce their consumption greatly — unless they’re eating tons of it.
We need better evidence, which means we need to shape up the field of nutritional research. This likely means halting further observational studies. We have years of data that led us to where we are now. As discussed in the editorial on these new studies, written by myself and a co-author, we have likely gained all the knowledge we can gain from them.
Everyone has heard the message that meat is unhealthy. Many people aren’t listening, and it’s not even clear they should. It’s time we offer up better evidence to a public that is hungry for the truth.